The path to mass market high precision positioning
Global navigation satellite system (GNSS) technology has come on in leaps and bounds in recent decades. Shortly after the turn of the century, for example, we saw the time it takes to achieve an accurate initial position reading slashed from minutes to less than 30 seconds. Later that decade, we saw receiver sensitivity improve, going from 130 dBm to 167 dBm. And where the American Global Positioning System (GPS) was the only worldwide positioning satellite constellation at the turn of the millennium, it’s since been joined by the European Galileo system, as well as China’s BeiDou and Russia’s GLONASS. And that’s before you add India’s NAVIC and Japan’s QZSS regional systems to the mix. This proliferation has enabled GNSS chip-makers to build receivers that work with multiple constellations. What’s more, the satellite signals have been modernized, and this year, for the first time, multi-band GNSS is becoming affordable. All of this provides the foundations for the next key GNSS focus topic: achieving accuracy down to the decimeter- or even centimeter-level.
The accuracy challenge
To identify its position, a GNSS receiver uses triangulation, picking up its distance from four or more satellites. This distance is calculated based on the time it takes for the satellite’s signal to get to the receiver. The difficulty is that an error of even a couple of billionths of a second can have a big impact on the accuracy of the reading. If there’s an error in the orbit position of the satellite, accuracy drops by up to 2.5 meters. A satellite clock error can lead to a further drop of up to 1.5 meters. Perturbations in the troposphere and ionosphere add further accuracy losses of 1 meter and 5 meters respectively – and more when the satellite is near the horizon, or when solar activity is especially intense. However, the biggest error occurs when signals from the satellite reach the receiver indirectly, or multiple times, as can happen when they bounce off building walls. This is known as the ‘multipath’ effect.
If you’re somewhere with open skies, a standard GNSS receiver will give you a reading with accuracy down to 2 meters.
Achieving high precision GNSS
GNSS correction data enables much higher levels of precision, by cancelling out GNSS errors. This correction data can be gathered using a base station in a known location, which monitors signals coming from the satellites and identifies any discrepancies between its actual position and the one obtained from the GNSS signals. These deviations are then sent to moving ‘rover’ vehicles that subscribe to the service, enabling them to pinpoint their position more accurately by using the correction data to adjust the reading from the GNSS satellites. If conditions are good, and the rover and base station are relatively close together, this technique can deliver centimeter-level accuracy.
However, while correction data can address satellite clock, satellite position and atmospheric errors, multipath errors are often unique to a device’s surroundings, meaning they may differ between the rover and the base station. As a result, these errors need to be tackled within the individual receiver device.
High precision GNSS has been around for some time, but because both the kit and correction services have been so expensive, it’s remained the preserve of a handful of specialist professions, such as surveying. But new technologies are now making high precision positioning accessible to the mass market. This is unlocking use cases such as lane-accurate navigation, precision aerial drone flights and landings, unmanned farm machinery, augmented reality and vehicle-to-everything (V2X) communication, where vehicles wirelessly communicate with one another and the road infrastructure to avoid collisions.
Delivering mass market high precision positioning using correction services
There are two approaches when it comes to delivering GNSS correction services. The first is observation space representation (OSR). Here, the service calculates the expected error at the location of each specific rover, and sends this directly to the rover device.
The other technique is called state space representation (SSR). Here, GNSS signal errors are monitored and then used to physically model errors spanning a full region, in a so-called ‘state space’ model. The data describing the model at a given point in time is transmitted to rovers right across the coverage area.
Only SSR can feasibly be scaled up to become a truly mass market solution. Here’s why.
Applicable in situations needing centimeter- or millimeter-level accuracy, OSR is used in real time kinematic (RTK) and network RTK satellite navigation. OSR-based systems need a two-way link between the rover and correction data service provider. Moreover, for optimal accuracy, the rover must remain within 30 km of the base station. The challenge with OSR is that if it were to be adopted by the mass market, current mobile communication networks would struggle to reliably deliver the levels of communication required. Consequently, OSR isn’t ideally suited to mass adoption.
SSR-based techniques, on the other hand, send out one stream of data that covers the full service area, and can be picked up by any rover. Thanks to this simpler communication approach, and the fact that you only need a reference station every 150-250 km, SSR is the only technique that can feasibly be used for mass market high precision positioning, including highly assisted driving.
Moreover, we’ll see even better performance as improved receiver hardware, able to pick up more data from satellites, is rolled out. Early GNSS satellites transmitted in just one frequency band. More modern ones use up to three: GPS, for example, sends out its signals in L1 (centered on 1575 MHz), L2 (1227 MHz) and L5 (1176 MHz). BeiDou and GLONASS both use L1 and L2. High precision receivers can benefit by using more than one frequency band from the same constellation, thereby significantly accelerating the speed at which they can achieve a highly precise reading. Ultimately, this results in a more robust and reliable location service.
High-precision GNSS systems of the future will be made up of a range of components working together. Firstly, you’ll have the existing satellite constellations. Second, there will be the reference base stations, logging the satellite signal errors in real time. Third are the correction services, which transmit the error components, via the Internet and geostationary satellites. And fourth will be the kit in the rover devices, including dual-band GNSS receivers, a cellular modem (to pick up correction data sent over the Internet), and an L-band receiver (to collect the correction data from the geostationary satellites).
Applying high precision positioning technology to the cars on our roads
Vehicles on our roads today are nearly all fully driver-controlled, but an ever-increasing proportion have some form of assistance system. As we shift towards completely autonomous driving, we’ll need to see the level of automation growing step-by-step in specific scenarios, such as parking or highway driving. In today’s assisted-driving vehicles, the human driver is still responsible for remaining in-lane or maneuvering between lanes (this is Level 1 in the diagram below). There are already vehicles that sit in the second level, with partially automated systems capable of holding or changing lanes in certain situations. As you move up the scale to Level 3, drivers will even be able to let go of the steering wheel in some scenarios, though they need to be ready to take back control if required. Step up to Level 4, and in some cases, you won’t even need a driver any more. Only once all this has been achieved, does the possibility of delivering fully driverless vehicles become the next feasible target (Level 5).
For autonomous driving to be safe, a range of technologies will need to work in harmony. It’s already possible to bring together camera images, radar and lidar information and high-definition maps to enable a vehicle to pinpoint its location down to 10 cm accuracy, and sense obstacles in a variety of situations. But on their own, these systems can’t fully replace a human driver. For example, as we move towards fully automated driving, a vehicle’s systems will need to rely on a precise position reading to decide whether it’s safe to switch from human to autonomous control. However, if surrounding conditions are poor or there are no distinguishing landmarks, the optical systems may not determine correctly whether autonomous mode can safely be activated. This is a particular issue in Level-4 systems, because it’s here that the driver can hand full control over to the autonomous systems in certain scenarios – deciding when this can be done safely is key.
In situations like this, high precision GNSS can be used with automotive dead reckoning. This brings together satellite navigation information with data from in-vehicle sensors, such as an accelerometer, gyroscope and wheel-speed detectors. This provides accurate and independent positional information, even when GNSS isn’t available. This precise position reading can help pinpoint which segment of a high definition map the vehicle is in, which, combined with geofencing of key areas, can be used to reduce speed for safety reasons, for example. The reading can also be used to calibrate the in-vehicle sensors.
ISO 26262 sets out the safety requirements for autonomous vehicles, including so-called ‘functional safety’. This covers the vehicle’s ability to safely respond to errors, either in its firmware or its hardware, ultimately to keep passengers safe. To meet these requirements, the sorts of systems we’ve just described are vital.
However, while functional safety is an essential component of autonomous vehicle safety, it only covers errors that could happen on the vehicle itself. As we’ve seen, when it comes to positioning, the key areas for error are external to the vehicle. So even if a vehicle was functionally safe, it wouldn’t know to reject inaccurate positional data. Consequently, you need a more comprehensive safety approach, called ‘integrity’. This would look at the complete technology landscape, including the sensors, security systems and V2X communications. Integrity demands that every piece of technology identifies how confident it is in the output it is providing. If this confidence drops too low, the vehicle will know when to switch to an alternative technology.
The foundations for future motoring
High GNSS accuracy is a central enabler of advanced driver-assistance systems (ADAS) and completely autonomous vehicles that ultimately improve road safety. Using multi-band receivers and SSR correction data, high precision GNSS provides a reliable reading of the vehicle’s position, irrespective of the circumstances. Accuracy will need to be down to the decimeter-level on open highways, and for now less than a meter on more complex city roads. And as well as being highly accurate when it comes to pinpointing the vehicle’s physical location, the position reading needs to be delivered with a very high degree of confidence in its accuracy. And lastly, if the technology is to achieve mass market adoption, it will need to work flawlessly and be affordable.